2025-2026 ENSO

[Stormchaserchuck1]

Atlantic (NAO) has slight inverse correlation from October to the Winter

Pacific has a pretty strong direct correlation from October to the Winter, a lot of that is around the PDO, which is strongest in Oct out of PNA/ENSO. 

[donsutherland1]

59 minutes ago, michsnowfreak said:

They are very biased towards enso and also biased towards a warmer climate. That is why periods of cold are never seen far out, you will see them grow colder as the timeframe nears. 

I suspect that its seeming inability to see cold anomalies except at shorter timescales has a lot to do with the idea that boundary conditions drive seasonal averages. ENSO, PDO are prominent conditions. The oceans overall are warming. Therefore, the model forecasts are tipped toward the warmer idea more broadly than is realistic. Worse, the coefficients of determination for such variables related to boundary conditions and actual seasonal outcomes are very low. 

These weak relationships reveal that other important factors are involved, including synoptic scale events that cannot be reliably forecast beyond 10-14 days. Some of these additional variables may not yet be known. Synoptic scale events i.e., large snowstorms, Arctic blasts, etc., can have a great influence on the overall seasonal outcomes. Thus, even a warm winter can be much snowier than normal or a cold winter can lack snowfall. On account of these other variables, every La Niña or El Niño event is not alike.

The seasonal models are not yet at a stage where they can even begin to consistently resolve the actual events that ultimately produce the seasonal outcome. A similar situation applies to subseasonal forecasting. Not surprisingly, beyond two weeks, model skill on the weekly guidance largely disappears. There also seems to be a larger deal of persistence in the two week or longer forecasts than what actually occurs.

AI may improve some of these outcomes. But even then, big challenges could still persist. For example, even as some experiments with random forest models have shown a degree of improved skill in forecasting ENSO, those models are constrained by their knowledge base. Hence, when it comes to forecasting extreme events e.g., super El Niño events, they have great difficulty.

Perhaps the combination of AI and quantum computing might produce some significant breakthroughs. But that's still in the future and perhaps a decade or more away, assuming society values science and basic research to make the investments necessary to arrive at that improved state of forecasting. That's an open question in some areas and it will become even more relevant as major states grapple with the costs of aging populations, rising debt relative to GDP, etc., and the trade-offs involved in making budget allocations.

[PhiEaglesfan712]

4 hours ago, roardog said:

I’ll take Nina over Nino every time around here in the winter. Hopefully we do have a “warm” October. It feels like a lot of times a cold October turns into a mild December. Although, that’s  probably due to a cold October being more common in a Nino. A cold October is kind of useless anyway except for chasing the first flakes of the season or some sloppy early season slush accumulation that melts in a few hours. 

Classic case is 2011. That was a cold October, with a snowstorm at the end of the month. Pattern flipped soon thereafter, and it was blowtorch from November through March.

[Mercurial]

9 hours ago, roardog said:

I’ll take Nina over Nino every time around here in the winter. Hopefully we do have a “warm” October. It feels like a lot of times a cold October turns into a mild December. Although, that’s  probably due to a cold October being more common in a Nino. A cold October is kind of useless anyway except for chasing the first flakes of the season or some sloppy early season slush accumulation that melts in a few hours. 

same

[snowman19]

The fact that the EURO is seeing it now speaks volumes. It’s finally showing a La Niña, the last to the party as always with its extreme ENSO warm bias

[snowman19]

Models have completely backed off on any Atlantic tropical threats for the next 2+ weeks. We are starting to approach the point where we are going to have to accept that this may end up being a below normal season (ACE/named storms) despite the La Niña

[bluewave]

23 hours ago, michsnowfreak said:

If most here follow model longterm seasonal snowfall forecasts over anything else, we have a problem . A longterm snowfall model forecast is probably the least accurate thing out there. Temp/precipitation bad enough....but snowfall? Model snow maps are to be taken with a grain of salt 48 hours out, much less 7 months out.

The reason the Euro seasonal snowfall forecast worked out from Philly to Boston is that the storm track matched the seasonal forecast. So it didn’t really matter for the snowfall outcomes that the overall forecast for DJF was too warm. Since the colder conditions than forecast last winter came in behind the warmer storm track.

[PhiEaglesfan712]

39 minutes ago, bluewave said:

The reason the Euro seasonal snowfall forecast worked out from Philly to Boston is that the storm track matched the seasonal forecast. So it didn’t really matter for the snowfall outcomes that the overall forecast for DJF was too warm. Since the colder conditions than forecast last winter all arrived behind the Great Lakes cutters after the warm storm tracks. 

I love how you cherry pick "Philly to Boston". Why not the mid-Atlantic (Philly to DC)? Oh wait, a lot of places there had a higher snow average the last 2 years than in 16-17 to 22-23, so it won't fit your agenda?

[GaWx]

On 9/5/2025 at 4:07 PM, mitchnick said:

Not that it matters at this point, but the monthly temp forecast isn't that bad for the NE. It's pretty rudimentary, but you get the jist of it that the 3 month averages of +1-2C is likely closer to +1 than +2.

 

1. Indeed. The Sept Euro fcast for the NE DJF anomaly is for an avg of only ~+1.8 F.

2. But that is almost as warm as any Sept fcast for the NE DJF of the last 9 with it just slightly colder than 24-5’s ~+2.0 F and near 22-3, 20-1, and 19-20’s +1.8 F.

3. How has the Sept Euro verified vs actual for La Niña in the NE? I’ll look at NYC:

For NYC using 1991-2020 avgs for the 5 Nina winters since 2017-8:

-24-5 ended up -1.4 F vs ~+2F Sept Euro fcast or ~3.4 F colder than Euro

-22-3 ended up +5F vs Euro ~+1.8F fcast or ~3.2F warmer than Euro

-21-2 ended up +1.1F vs Euro ~+0.4F fcast or ~0.7F warmer than Euro

-20-1 ended up 0.0F vs Euro +1.8F fcast or 1.8F colder than Euro

-17-8 ended up +0.2F vs Euro +0.9F or 0.7F colder than Euro

 So, for the 5 Niña winters since 17-8, the Sept Euro turned out to be in F:

3.4 too warm, 3.2 too cold, 0.7 too cold, 1.8 too warm, and 0.7 too warm or an avg of 0.4 too warm

[snowman19]

13 hours ago, PhiEaglesfan712 said:

Classic case is 2011. That was a cold October, with a snowstorm at the end of the month. Pattern flipped soon thereafter, and it was blowtorch from November through March.

2011-12 didn’t turn into an all out disaster until the end of November. A few days after Thanksgiving that massive Bering Sea vortex developed and just sat and spun there for months on end…right through April. It was one of the most persistent patterns I’ve ever seen. There was damn near close to consensus on east having another cold and snowy winter and all those forecasts busted horribly 

[bluewave]

1 hour ago, PhiEaglesfan712 said:

I love how you cherry pick "Philly to Boston". Why not the mid-Atlantic (Philly to DC)? Oh wait, a lot of places there had a higher snow average the last 2 years than in 16-17 to 22-23, so it won't fit your agenda?

DC used to average more than 20” of snow a year. So their snowfall has been in steady decline for years now. The much lower bar allowed DC to reach near average this year at 14.9”.

DC missed out on the snowy 16-17 and 17-18 seasons further north. So this is the first 9 year stretch there with under a 10” average.

If your average snowfall eventually gets low enough, then it can be easier to reach the new lower average with one storm. That’s why the further south you get in the U.S. the easier it becomes for one storm to get you closer to the new lower snowfall averages. 

In a cutter, hugger, and suppressed Southern Stream storm track regime for Philly to Boston, sometimes the DC area can get a snow event which gets suppressed south of the Philly to Boston I-95 corridor. 

Makes sense that the long term snowfall decline is more pronounced around DC since they are further south where it’s warmer with more  marginal temperatures to begin with. 
 

 

 

 

[donsutherland1]

2 hours ago, snowman19 said:

Models have completely backed off on any Atlantic tropical threats for the next 2+ weeks. We are starting to approach the point where we are going to have to accept that this may end up being a below normal season (ACE/named storms) despite the La Niña

What this means is that the social media hurricane hypesters will simply shift the goalposts to a few weeks down the road to maybe late September or October (an art that many have perfected from punting long-range cold/snow calls when they don't verify). I suspect that there will be some activity before the season concludes, so they will also proclaim that they were correct. Also, there's no reason the current season can't end with only a modest uptick in activity. Every season does not end with a spectacular burst of late-season storms.

There's a window currently suggested on the MJO guidance for the second half of September (probably after the 20th) into the first week of October. Afterward, things could largely shut down. The forecast lift isn't as great or widespread as one would look for if there were to be a big explosion of tropical cyclone genesis or a large number of hurricanes.

[snowman19]

@Bluewave Once again, tropical forcing staying in the IO and the Maritime Continent regions…..
 

 

 

 

 

[bluewave]

25 minutes ago, snowman19 said:

@Bluewave Once again, tropical forcing staying in the IO and the Maritime Continent regions…..
 

 

 

 

 

 

 

[roardog]

3 hours ago, snowman19 said:

2011-12 didn’t turn into an all out disaster until the end of November. A few days after Thanksgiving that massive Bering Sea vortex developed and just sat and spun there for months on end…right through April. It was one of the most persistent patterns I’ve ever seen. There was damn near close to consensus on east having another cold and snowy winter and all those forecasts busted horribly 

If that happens this winter, I expect if someone working in an Antarctic research station posts a tweet about it,  you’ll find it and post it here. lol

[bluewave]

On 9/6/2025 at 10:46 AM, 40/70 Benchmark said:

Did NOT get the pattern right, though.....suprised someone as meticulous as you would omit that.

If a seasonal forecast model gets any aspect of its winter forecast correct, then I will give it credit. The storm track aspect and well below normal snowfall forecast worked out. Even if the temperatures and 500mb pattern were significantly off.

So the very fast Pacific flow and Southeast ridge found ways to emerge right during our biggest storm dates last winter.

The last time we saw any seasonal forecast get most of its seasonal forecast correct may have been the JMA for the 13-14 winter issued during that October.

As a general rule, computer modeling doesn’t have much skill beyond 1 to 2 weeks out.

My guess is that they just run with a stock composite for whichever phase of ENSO we are in. So when we got a mismatch like my early outlook last October suggested, the stock La Niña composite didn’t work out. 

The record WPAC warm pool prevented the Euro colder pattern forecast for the 23-24 El Niño in much of the East from developing. Instead we got the record warmth.

During the 22-23 La Niña the much stronger -PDO interaction with that pattern resulted in a much deeper Western trough than forecast. So we got the record warmth in the East that the model missed. 

21-22 had the record December -PNA and warmth which the seasonal forecast couldn’t see. Plus the MJO 8 in January which couldn’t be seen beyond a few weeks out. The overall 3 month average for that winter wasn’t too far off. But of course the finer details were significantly different.

20-21 missed the +PNA December mismatch again from the October MJO signal that year. It was the first time I used it for a seasonal discussion that year back in the NYC Metro forum.

The 19-20 colder Euro forecast was off that year since the El Niño couldn’t couple due to the record WPAC warm pool along with the fall IO forcing leading to the record SPV and strong +AO +NAO.

The Euro 18-19 forecast was also off since its colder El Niño forecast couldn’t couple due to the record WPAC warm pool. 

 

 

 

[40/70 Benchmark]

2 hours ago, bluewave said:

If a seasonal forecast model gets any aspect of its winter forecast correct, then I will give it credit. The storm track aspect and well below normal snowfall forecast worked out. Even if the temperatures and 500mb pattern were significantly off.

So the very fast Pacific flow and Southeast ridge found ways to emerge right during our biggest storm dates last winter.

The last time we saw any seasonal forecast get most of its seasonal forecast correct may have been the JMA for the 13-14 winter issued during that October.

As a general rule, computer modeling doesn’t have much skill beyond 1 to 2 weeks out.

My guess is that they just run with a stock composite for whichever phase of ENSO we are in. So when we got a mismatch like my early outlook last October suggested, the stock La Niña composite didn’t work out. 

The record WPAC warm pool prevented the Euro colder pattern forecast for the 23-24 El Niño in much of the East from developing. Instead we got the record warmth.

During the 22-23 La Niña the much stronger -PDO interaction with that pattern resulted in a much deeper Western trough than forecast. So we got the record warmth in the East that the model missed. 

21-22 had the record December -PNA and warmth which the seasonal forecast couldn’t see. Plus the MJO 8 in January which couldn’t be seen beyond a few weeks out. The overall 3 month average for that winter wasn’t too far off. But of course the finer details were significantly different.

20-21 missed the +PNA December mismatch again from the October MJO signal that year. It was the first time I used it for a seasonal discussion that year back in the NYC Metro forum.

The 19-20 colder Euro forecast was off that year since the El Niño couldn’t couple due to the record WPAC warm pool along with the fall IO forcing leading to the record SPV and strong +AO +NAO.

The Euro 18-19 forecast was also off since its colder El Niño forecast couldn’t couple due to the record WPAC warm pool. 

 

 

 

Well, that is where we differ. I know in the past I have ostensibly nailed seasons, but always detract from the grade when I'm "right for the wrong reason".

Agree to disagree. The forecast was wrong....it just so happen to be correct regarding low snowfall.

[bluewave]

28 minutes ago, 40/70 Benchmark said:

Well, that is where we differ. I know in the past I have ostensibly nailed seasons, but always detract from the grade when I'm "right for the wrong reason".

Agree to disagree. The forecast was wrong....it just so happen to be correct regarding low snowfall.

There are really three key elements of a seasonal forecast. General seasonal temperatures, storm tracks which includes P-types  and amounts, and the 500 mb patterns. 

The Euro missed on the temperatures and 500mb heights. But the storm tracks and snowfall amounts were mostly right. 

So it just could be that the model stock La Niña forecast only worked out for the storm track through the Eastern Great Lakes. Unfortunately, this has been the default primary storm track since 18-19. 

 

 

[GaWx]

1 hour ago, 40/70 Benchmark said:

Well, that is where we differ. I know in the past I have ostensibly nailed seasons, but always detract from the grade when I'm "right for the wrong reason".

Agree to disagree. The forecast was wrong....it just so happen to be correct regarding low snowfall.

Ray, I graded the Sept ‘24 Euro 2M temp. forecast for DJF 2024-5 an F, the worst Sept winter forecast going back at least to that of Sept of 2017, the furthest back I can find. The Euro averaged ~4F too warm for most of the US! The only area it did well with was the SW. 

 The Sept ‘20 Euro forecast for 20-21 was similarly too warm but not by quite as much (avg of ~3F too warm for the US).

 OTOH, the Sept forecast for 2023-4 was pretty bad in the other direction with an average miss of ~3F too cold for the US overall.

 The last good Sept Euro 2M forecasts for the lower 48 as a whole were for 2021-2 and 2019-20. The avg miss in the NE was very small.

[GaWx]

6 hours ago, snowman19 said:

@Bluewave Once again, tropical forcing staying in the IO and the Maritime Continent regions…..
 

Snowman, the GFS and EPS MJO forecasts (the two best imho) for the next 2 weeks agree that the MJO will remain in the IO and then reverse back to the W Hem/Africa. There’s no MC in sight:

 

[bluewave]

25 minutes ago, GaWx said:

Ray, I graded the Sept ‘24 Euro 2M temp. forecast for DJF 2024-5 an F, the worst Sept winter forecast going back at least to that of Sept of 2017, the furthest back I can find. The Euro US averaged ~4F too warm for most of the US! The only area it did well with was the SW. 

 The Sept ‘20 Euro forecast for 20-21 was similarly too warm but not by quite as much (avg of ~3F too warm for the US).

 OTOH, the Sept forecast for 2023-4 was pretty bad in the other direction with an average miss of ~3F too cold for the US overall.

 The last good Sept Euro 2M forecasts for the lower 48 as a whole were for 2021-2 and 2019-20. The avg miss in the NE was very small.

If we are looking for as close to perfection as we can get from an October seasonal forecast, then the JMA October 2013 record TNH forecast was pretty close.
 

https://ds.data.jma.go.jp/wmc/products/model/map/7mE/map1/pztmap.php

 

[40/70 Benchmark]

37 minutes ago, bluewave said:

There are really three key elements of a seasonal forecast. General seasonal temperatures, storm tracks which includes P-types  and amounts, and the 500 mb patterns. 

The Euro missed on the temperatures and 500mb heights. But the storm tracks and snowfall amounts were mostly right. 

So it just could be that the model stock La Niña forecast only worked out for the storm track through the Eastern Great Lakes. Unfortunately, this has been the default primary storm track since 18-19. 

 

 

Yea, I'm not arguing any of that...but temps and 500mb heights are kind of a big deal. I mean, I nailed snowfall last season...but I was too warm....that can't be ignored.

[bluewave]

7 minutes ago, 40/70 Benchmark said:

Yea, I'm not arguing any of that...but temps and 500mb heights are kind of a big deal. I mean, I nailed snowfall last season...but I was too warm....that can't be ignored.

When I was a kid growing up, we were lucky if even the next few days forecast was correct. The computer models were really primitive during the 70s and 80s.

But the Euro doing so well with March 93 and January 96 events began to give us some confidence in the 3-5 day forecast. Occasionally the model does really well on storm details day 6-10 like with Sandy.

But the range beyond week 1 and week 2 is still very low skill. Maybe they can find a way for AI to bias correct the longer range NWP forecasts. 

[GaWx]

 The OHC has recently been cooling substantially after the prior rather steady period:

@snowman19you’ll probably like this trend